Alibaba’s HappyHorse 1.1 rockets to No. 2 as Sora and Seedance implode for enterprises

Alibaba Cloud says its AI video model is now production-ready, and the benchmark crowd just moved it up the board fast. HappyHorse 1.1 is live on Alibaba Cloud Model Studio with full API access for enterprise customers and developers, plus a 40% sitewide launch discount for the first two weeks. The timing matters because this is not happening in a stable market, it is happening while competitors stumble out of the room.

The interruption is real and unusually sharp. OpenAI discontinued Sora on April 26, and the Sora API is set to follow on September 24. ByteDance indefinitely postponed the international rollout of Seedance 2.0 after major Hollywood studios sent legal threats, alleging systematic copyright infringement. For enterprise teams evaluating or integrating AI video tools into marketing, advertising, and content production workflows, the competitive landscape has contracted sharply in a matter of months. Alibaba is essentially walking into a narrower marketplace with a new API-first product pitch and a serious distribution engine behind it.

So what did Alibaba actually ship? HappyHorse 1.1 is a major upgrade to HappyHorse, which first appeared in early April as an anonymous submission on the Artificial Analysis Video Arena, where real users compare model outputs in blind, side-by-side evaluations. The model immediately claimed the top position in both text-to-video and image-to-video rankings. Arena.ai later confirmed the creator as Alibaba, revealing the work was built by the company's ATH (Alibaba Token Hub) AI Innovation Unit, previously part of the Future Life Lab under the Taobao and Tmall Group before a strategic organizational restructuring.

At the same time, Arena.ai’s tracking shows HappyHorse 1.0 now holds the No. 2 position across all three Video Arena leaderboards. Arena noted the model scores 1,444 in both text-to-video and image-to-video categories. In those comparisons, it leads Google's Veo-3.1 (with audio) by 69 points in text-to-video and xAI's Grok-Imagine-Video by 23 points in image-to-video. In Elo-based ranking systems like Arena's, model points move based on user preference in head-to-head tests, meaning persistent double-digit leads reflect a consistent quality gap as perceived by human evaluators, not a random statistical wobble.

HappyHorse’s architecture is part of why it matters for enterprise deployments, not just benchmark bragging. Community-compiled technical documentation describes a 15-billion-parameter unified self-attention Transformer that processes text, image, video, and audio tokens within a single token sequence. Unlike many competitors that stitch separate models for video and audio together, HappyHorse operates as a unified system that handles all modalities in a single generation pass. That reduces the need for third-party dubbing or post-processing audio tools, which is a big deal when you are trying to control total cost of ownership and time to production. Fewer integration points and fewer vendor dependencies tend to translate into fewer surprises in procurement, security review, and production pipelines.

The 1.1 upgrade is where Alibaba leans into the pain points that usually slow down adoption. Alibaba describes HappyHorse 1.1 as “systematically optimized across core content generation scenarios,” and the specific improvements map to issues enterprise teams run into repeatedly. The most consequential is multi-image reference capability, called R2V (Reference-to-Video). Users can upload multiple character reference images and maintain consistent identity across generated video. For brands producing advertising campaigns, product videos, or serialized marketing content, keeping a character’s look stable across shots is not a “nice-to-have.” It is often the difference between shipping a campaign and scrapping a batch.

Alibaba also highlights strengthened motion modeling aimed at speed and fluidity, plus targeted visual texture improvements. It specifically calls out the elimination of “facial oiliness,” “over-sharpening,” and “unnatural textures,” artifacts that have historically signaled machine-generated content to viewers. On the audio side, HappyHorse 1.1 improves audio-visual synchronization, including what Alibaba claims is “zero-drift lip sync” for dialogue scenes and context-aware speech pacing. It also improves instruction-following for long and complex prompts, which matters when an enterprise needs precise camera movements, lighting conditions, and narrative beats in one generation pass rather than iterating through dozens of attempts.

Why is this launch arriving with such upside for Alibaba? Because two major rivals have hit different kinds of walls, leaving fewer options for buyers. OpenAI discontinued Sora’s web and app experiences on April 26, while the Sora API is planned to end on September 24. The source also frames the business case for that decision in blunt terms: Sora cost roughly $1 million per day to operate but generated only about $2.1 million in total revenue. Active users dropped from a peak near 1 million to under 500,000. Meanwhile, ByteDance’s Seedance 2.0 faced allegations of systematic copyright infringement. Legal threats came from Netflix, Warner Bros., Disney, Paramount, and Sony after users generated viral clips featuring Hollywood intellectual property. ByteDance indefinitely postponed the international launch, and the global rollout remains suspended.

In that narrower landscape, the primary Western competitor mentioned is Google’s Veo 3.1. Arena rankings suggest HappyHorse is outperforming Veo on user-perceived quality. But rankings alone do not close enterprise deals. Alibaba also sweetened the offer with pricing. At the 1.0 level, pricing through third-party API platforms ran roughly $1.82 per 10-second clip at 720p and $3.12 at 1080p. With the promotional pricing, HappyHorse 1.1 could make production-quality AI video generation feasible for mid-market companies and agencies that previously treated the tech as too expensive for anything beyond experimentation.

Zoom out further and Alibaba’s push is not just a model release. The source says HappyHorse 1.1 sits atop a global infrastructure buildout, citing a $52.7 billion infrastructure bet. That matters because it gives Alibaba distribution advantages that pure-play AI model companies may struggle to match. For enterprise buyers, this can reduce the friction of running AI video workloads at scale, while for executives and boards it changes the risk math. If the market’s top tools are disappearing or pausing, the enterprise question stops being “who has the best demo” and becomes “who can reliably deliver production workflows at scale.”

That is the test Alibaba is taking on right now: turning benchmark momentum into enterprise adoption, including in Western markets navigating intensifying U.S.-China tech tensions. If HappyHorse 1.1 succeeds, it will not just fill a technical gap. It will likely define what “enterprise-ready” generative video looks like when the competitive field is actively shrinking.